The data were assigned an error floor of 10% (2.9°) and 5% (1.45°), for phases 15 and 10%, for the TE- and TM-mode resistivities, respectively.
Moreover, in the training phase, 70% of the data is assigned for training, 15% for testing, and 15% for validation (Wang et al. 2002).
In this step, the real output of each data is compared with the outputs of the estimated subsystems and then the data is assigned to a region with least error.
The data were assigned to two groups (OA and old) and DA was performed.
STRUCTURE version 2.2 calculates a logarithmic probability for the data being assigned to a given number of clusters, based on minimizing linkage between clusters, and maximizing linkage within.
The regularization parameter, which allows the degree of shrinkage to be estimated from the data, was assigned a prior Beta distribution as outlined in [ 37], with parameters shape = 0.52 and rate = 1 × 10−4 yielding λ = 57.5 for our data.
If not, a rank transformation was applied: that is, the data were assigned ranks, as for a non-parametric analysis, but the ranks were then subjected to parametric tests.
2) Nonparametric empirical Bayes correction normalization (NEB) [ 14]: The drawback of the linear normalization is that the proportion p of the missing reads (because the sequencing depth is not enough, some reads would be missing) from the data is assigned to be zero, which underestimates the p of highly abundant reads and overestimates the p of low and intermediate abundance reads.
In the studies where data are available only at a single location in the distribution system, these data are assigned directly to the subject's residential tap water.
